It's possible if you have the sample size and a test statistic (t-value, p-value) of the estimate.

Hello Shaun,

If you know: (a) R-squared for the model; (b) number of IVs (k) for the model; (c) N of cases; and (d) SD of the dependent variable, Y:

SSE = (1 - R^2) * SD^2 * (N - 1); (sum of squares error/residuals)

VEE = SSE / (N - k - 1); (variance error of estimate)

SEE = sqrt(VEE) (standard error of estimate)

Good luck with your work.

Note: Edited to fix omission in SSE equation. Happens when you get in a hurr.

Hello Shaun Joffe. Do any of the ones that fail to report the SE of Estimate report any of the following?

• Root Mean Square Error (RMSE)
• MSerror
• MSresidual

Apologies if you already know that RMSE = the SE of estimate. It's always a guessing game on here trying to work out what posters do and do not know. ;-)

Shaun -

Please note that often this measure would not mean much due to heteroscedasticity. (See https://www.researchgate.net/publication/354854317_WHEN_WOULD_HETEROSCEDASTICITY_IN_REGRESSION_OCCUR.)

With larger predicted-y you would often have larger sigma for the residuals. Thus the same regression model would typically have a different such measure if applied over a different set of predicted-y cases.

Best wishes - Jim

Provide output, and say sample size.